In short, the lower numbers the better (other factors being equal). Less space consumed on the disk, less RAM on the Graphic card, and more efficient rendering.
In openBRF, you can achieve the same result if you set the option "after import" to "merge". Try.
To really answer your question, it takes to have a look at the internal structure of BRF meshes...
In M&B, a mesh is composed of
- an array (a set, a group, a bunch) of triangles [first number]
- an array of vertices [second number]
- an array of XYZ positions [third number]
But, wait, aren't the last two the same thing?
No. A vertex has normals and uv text coords, but no XYZ coordiantes. Instead, it has an index to a position, and the position has XYZ coords.
Why so? because that way two "vertices" which are in the same position but have different texture coords or normals
(this is called a seam), they can point at the same "position", and the actual XYZ coordinates are not repeated for the second vertex.
So the mapping between vertices and positions is not 1:1 and there can be more vertices than positions.
Summary:
- a face has only three indices, to three different vertices
- a vertex has: u,v texture coords, normals and an index to a position
- a position has XYZ coordintes, and rigigng (weitght and bone indices, for up to 4 bones).
But mesh formats (e.g. obj) are not like that. They have only faces and vertices. Each vertex has X,Y,Z coords inside it, directly.
So, two vertices in a seam (i.e. same position, different uv coords) they repeat their coordiantes, each has its own copy of XYZ.
This is commonly called an indexed mesh.
(SMD files has even less: they have only triangles: normals, u,v coords, rigging etc is repeated three times for the three vertices
of each face.)
Therefore, after loading, there is the opportunity to save space/resources by merging together positions that are identical but that were separated in the original (obj), and even vertices (in SMD files)... This is what the "unify" option does. If you also ignore the original normals, by recomputing them, you can unify more veritces because less vertices will be distinguished on the basis that they originally had a different normals (like in the flat shaded giant above). This is what the "unify and recompute normals" option does.
Thanks for asking. I now see that the default setting should be to "unify" meshes after import. The only reason not to unify is to keep vertex order unaltered, to merge frames in an animation (in the future).
(I saved during edits because this browser crashes a lot on me)
Two vertices will be merged if they are totally identical: same u,v, coords, same normal, and they refer to the same "position" (so they have same XYZ coords and same rigging).
Two positions will be merged if they have the same XYZ coordiantes, and the same rigging.